
Full-Stack Engineer (AI & LLM Integration)
JetBridge
Posted about 2 hours ago
They have a rapidly built, highly functional frontend prototype generated via low-code tooling (React/Next.js) that successfully validates about 70% of the product vision. As they transition from validation to scale, they need a Full Stack Engineer to lead the migration from a prototype architecture to a production-ready system. This includes refactoring the frontend, designing a robust backend architecture from scratch, and implementing a formal orchestration layer for our LLM and image generation pipelines.
Responsibilities:
-
Clean up, refactor, and modularize a rapid-prototype React/Next.js frontend. Design and implement a robust, production-ready backend architecture from scratch.
-
Move direct LLM API calls into a structured orchestration layer. Build and optimize advanced prompt engineering workflows and image generation pipelines (specifically leveraging OpenAI's image generation models/DALL-E).
-
Set up data ingestion workflows to parse clinical study protocols. Prepare and structure internal ad-performance training data for potential LLM fine-tuning.
-
Own the deployment, hosting, and monitoring strategy to transition the app from a fragile prototype to a stable, scalable MVP.
-
Ensure best practices for data handling (Note: the app processes study protocols and marketing data, not patient data/PHI, so full HIPAA/SOC 2 compliance is not required for this phase).
-
4+ years of professional full-stack development experience (comfortably operating at a strong Mid to Senior level).
-
Deep expertise in React and Next.js.
-
Experience building production-grade backends (Node.js) and deploying apps to stable cloud environments.
-
Hands-on experience integrating LLMs, managing image generation pipelines, handling API rate limits, and implementing orchestration frameworks (e.g., LangChain, LlamaIndex, or custom structured outputs).
-
A "scrappy but disciplined" engineer who isn't afraid to dig into messy, auto-generated code and systematically refactor it without losing momentum.



